AInstAIn LogoAInstAIn, a free AI agent with live web search

How it reasons, which open-source models it uses and how to use it: the technical guide

AInstAIn is an AI agent: a system that does not just answer, but reasons in several steps, looks up up-to-date information on the web, uses tools and compares several language models before giving you the answer. It is built on open-source and open-weight LLMs such as GPT-OSS and Qwen, run on Groq infrastructure, and is designed and run in Italy by VinDimax.

Try AInstAIn for free

What is AInstAIn

AInstAIn is an AI agent you can use in the browser and install as an app on your phone (PWA). Unlike a traditional chatbot, which generates text in a single step, an agent breaks the request down, picks the right tools, checks the results and only then replies.

The name is written AInstAIn, but you may also see it as Ainstain or AI Instain: it is the same AI agent, at ainstain.site.

It works in eight languages (Italian, English, French, German, Spanish, Chinese, Japanese, Russian) and detects the language you write in. It offers three personalities that change the style of the answers, not their quality: AInstAIn (neutral and precise), Elia (direct and pragmatic) and Sofia (empathetic and clear).

How an AI agent reasons

In Agent Mode the system applies a structured reasoning loop, known as the ReAct pattern (Reason + Act):

Reasoning loop of an AI agent The question enters the loop: thought, action with a tool, observation of the result. The loop repeats up to eight steps, then the answer comes out with its sources. Question 1. Thoughtwhat do I need? 2. ActionI pick a tool 3. ObservationI read the result Enough?no: repeat (max 8) Answerwith sources tools: web search · date and time ·calculator · memory · images
The thought → action → observation loop repeats up to a maximum of eight steps.
  1. Thought. The model analyses the request and decides which piece of information is missing.
  2. Action. It picks a tool: web search, reading the date and time, a maths calculation, saving to memory or generating an image.
  3. Observation. It reads the tool's result and compares it with the goal.
  4. Answer. When it has enough data, it writes the final answer and cites its sources; otherwise it repeats the loop, up to eight steps.

Up-to-date data: the source rule

For news, sports results, prices, weather and any data that changes over time, AInstAIn runs a web search and builds the answer only on the results it finds, naming the source. If the data does not emerge from the search, it says so instead of making it up. Web search is available to all users, including the free plan.

Multi-AI answer

In Multi-AI mode several models answer the same question (Qwen 3.8, GPT-OSS 120B and GPT-OSS 20B) and a synthesis model compares the answers, checks their consistency and writes the final one. There is also a fast mode, which favours speed.

Service continuity

If the main provider does not respond, the request automatically switches to a fallback provider (OpenRouter, with free models), so the service stays available.

Built on open-source models

AInstAIn does not depend on a single proprietary model: it uses open-source and open-weight models run on Groq, a very fast inference platform with a free tier. This makes it possible to offer a free service without giving up speed.

GPT-OSS 120B

OpenAI's open-weight model, used as the main model for chat and reasoning.

GPT-OSS 20B

A lighter, faster version, used for quick answers and as the synthesis model.

Qwen 3.8

Alibaba's open-weight multimodal model: it works alongside the others in Multi-AI and in image analysis.

The available models change over time: AInstAIn updates them when providers release new ones. You can compare chatbots on the Ratings page (in Italian), where each assistant is judged on 42 categories.

What makes AInstAIn different

An agent, not just a chatbot

It reasons in several steps and uses tools: web search, calculations, images, code.

Several models, one answer

Multi-AI compares different models and synthesises the best answer.

Cited sources

For up-to-date data it uses only verifiable web results and says where they come from.

Open source and free

Built on open-source LLMs and infrastructure with a free tier: the chat is free to use.

Three personalities, eight languages

AInstAIn, Elia and Sofia; the language automatically follows your message.

No account needed

Start chatting right away: your chat history stays in your browser.

What is free and what needs the code

FeatureFreePremium
Chat with 3 personalities and 8 languages✓✓
Real-time web search with sources✓✓
Agent Mode (multi-step reasoning)—✓
Multi-AI answer—✓
Documents (PDF, Word, Excel, PowerPoint…)—✓
Image analysis and generation—✓
Coding Mode with live preview—✓
Voice input—✓

The Premium plan is activated with an access code, entered from the plan menu.

How to use it

  1. Open the chat. Go to the AInstAIn home: no sign-up needed.
  2. Choose a personality. AInstAIn, Elia or Sofia: hover over the name to see the style.
  3. Write in your language. For example: "What is the latest news on the Champions League?". The page and the answer follow the language of your message.
  4. Turn on the tools (Premium). From the Tools menu you can switch on Agent Mode and Multi-AI, and from the chat bar attach documents or images.
  5. Check the sources. Answers with up-to-date data show a "web-updated answer" label and cite the source.

Areas for improvement

AInstAIn is under continuous development and uses free models: here is where we are working.

Frequently asked questions

What is an AI agent?

An AI agent is a system that does more than answer: it breaks a request into steps, chooses the tools to use (for example web search or a calculation), looks at the results and keeps going until it has a complete answer. A traditional chatbot, by contrast, produces its answer in a single step from what the model has already learned.

What is the difference between an AI agent and a chatbot?

A chatbot generates text in one step. An agent repeats a reasoning loop (thought, action, observation) and can use external tools, so it can solve tasks that need up-to-date data, calculations or several checks in sequence.

Is AInstAIn free?

Yes, there is a Free plan: it includes chat with three personalities, eight languages and web search for up-to-date answers. The Premium plan, activated with an access code, unlocks Agent Mode, Multi-AI answers, document upload, image analysis and generation, Coding Mode and voice input.

Which models does AInstAIn use?

AInstAIn uses open-source and open-weight models: GPT-OSS 120B and GPT-OSS 20B (OpenAI) and Qwen 3.8 (Alibaba), run on Groq infrastructure. If the main provider does not respond, it switches to a fallback on OpenRouter. The list of models is updated when providers release new ones.

How does it decide whether to search the web?

For news, sports results, prices, weather and any data that changes over time, the system runs a web search and uses only the results it finds, citing the source. If it cannot find the data, it says so instead of making it up.

Is my data safe?

No account is needed and your chat history stays in your browser. To generate a reply, the text of your message is sent to the AInstAIn service and to the model providers (Groq and, as a fallback, OpenRouter): do not enter confidential or sensitive data.

Does AInstAIn work in other languages?

Yes. The interface is available in Italian, English, French, German, Spanish, Chinese, Japanese and Russian, and the reply automatically follows the language you write in.

Learn more

Try AInstAIn for free